<>1. Mediating effect
Mediating effect model ：A Yes C The role of B happen , Namely A-B-C. among A-C If the effect is zero , be B It is a complete intermediary ; if A-C The effect is not zero , be B Part of the intermediary .
Figurative metaphor ： The mediating effect was as follows “ woman matchmaker ”,A-C The understanding is through matchmaker .
<>1.1 Mediating effect
It refers to the influence relationship between variables （X→Y） It's not a direct causal chain , But through one or more variables M Indirect impact , At this time M Is the intermediary variable ,X adopt M Yes Y The indirect effect is called mediating effect .
The above basic model and regression equation describe the relationship between variables ：
equation （1） Coefficient of c by X Yes Y The total effect of ;
equation （2） Coefficient of a yes X Yes M The direct effect of ;
equation （3） Coefficient of b It's in control X After the impact of ,M Yes Y The direct effect of ;
coefficient c’ It's control M After the impact of , X Yes Y The direct effect of ; coefficient ab Through the mediation variable M The mediating effect produced , And exist ab=c-c’ The relationship between .
<>1.2 Analysis steps
Step 1 ：X Yes Y The return of , Test regression coefficient c The significance of
Step 2 ：X Yes M The return of , Test regression coefficient a The significance of
Step 3 ：X and M Yes Y The return of , Test regression coefficient b and c’ The significance of
stay SPSS Operating software , For each equation (1)(2)(3) Linear regression analysis was performed , The significance of coefficient was tested step by step . Open menu bar , analysis → regression → Line type , Independent variable and dependent variable were added respectively , Output results , The significance of the coefficient is obtained .
1.31 Data display
To study the psychological factors between job identity and job performance （ anxious ） The meaning of , Case data include “ Not recognized ”,“ anxious ”,“ job performance ” 3 Variables .
Dependent variable , dependent variable , To understand the concept of intermediary variables ,“ Not recognized ” The independent variable X,“ anxious ” The intermediate variable M,“ job performance ” The dependent variable Y.
1.32 Operation process
Step1： Test equation Y=c*X+e1 Medium coefficient c Is it significant
The operation is very simple , It's a conventional linear regression process . menu ：【 analysis 】→【 regression 】→【 linear 】, You can operate in the main dialog box of linear regression .
Linear fitting results ：
obviously , Model Y=c*X+e1 remarkable , Standardization coefficient c=0.678,p=0.000, remarkable . We can continue to test whether the other two equations are significant .
Step2： Test equation M=a*X+e2 Medium coefficient a Is it significant
The linear regression process was repeated , Anxiety variable as dependent variable , The work is not recognized variables as independent variables for linear fitting .
obviously , Model M=a*X+e2 remarkable , Standardization coefficient a=0.533,p=0.000, The coefficient is significant . We can continue to test another equation .
Step3： Test equation Y=c’X+bM+e3 Medium coefficient b and c’ Is it significant
The linear regression process was repeated , Job performance as dependent variable , Job dissatisfaction and anxiety are both independent variables , Linear fitting is enough .
obviously , Model Y=c’X+bM+e3 remarkable , Standardization coefficient b=0.213,p=0.000, The coefficient is significant . coefficient c’=0.564,p=0.000, remarkable .
In this case, the equation (2) equation (3) Related to “ anxious ” Coefficient of variable a Sum coefficient b All were significant , equation (3) in c’ remarkable , Therefore, this study belongs to partial mediating effect .
independent variable “ Work is not recognized ” For dependent variable “ job performance ” The mediating effect is not completely through the mediating variable “ anxious ” To achieve the impact ,“ Work is not recognized ” Yes “ job performance ” There are some direct effects .
Contribution rate of mediating effect to total effect ：M=ab/c=0.5330.213/0.678=0.167, and 16.7%.
<>2. Regulatory effect
Regulatory effect ：A-C It works , but B Will affect A-C The effect of .
Figurative metaphor ： The regulatory effect is the regulatory effect “ the other woman ”, Will affect A-C A normal relationship .
The picture above shows i Model of interaction ：A-C It matters ,B-C It matters ; also B Will affect A-C relationship ,A Will affect B-C relationship .
This is like A and B It's a roommate in the same dorm , I like them all at the same time C, meaning AB Each other is a junior , But there is no priority .
There is no essential difference between regulatory effect and interaction effect in statistical model ; But the moderating effect can specify who is the independent variable , Who is the moderator ; The interaction status is equivalent .
Study the mediating and regulatory effects , However, when the research factor is a significant variable, it is used Process optimum ;
When it is a latent variable, it is used AMOS For the better , And of course lisrel,Mplus etc .
lisrel It is the earliest structural equation modeling software , Operation by programming , Has been gradually replaced , Those who are interested can study hard Process and AMOS bar !